A multispectral imaging system and method are disclosed for extracting useful scene information from spatial pictures of an image scene in many different spectral wavelength bands. A real time optical processor is used to perform optical correlation.
Interest in multispectral imaging systems is rapidly increasing. Such systems record spatial pictures of an image scene in many different spectral wavelength bands, e.g., a scene image at red wavelength, green wavelength and is blue wavelength. Differences in the observed spatial image at different wavelengths has been shown to be very useful for finding hidden targets, assaying agricultural conditions, and detecting other subtle features that would not be noticed in a video picture of the scene. Such systems have evolved from several spectral bands, referred to as multispectral, to hundreds of bands, known as hyperspectral, and will continue to evolve into even finer spectral xe2x80x9cbinningxe2x80x9d, known as ultraspectral. As this evolution has progressed, the amount of data generated by a multispectral imager has grown to the point where it frequently exceeds the capacity to data link it to a remote location for the extensive of amount of signal processing necessary to extract useful information.
This has led to substantial activity to develop xe2x80x9creal-timexe2x80x9d processors located at the instrument. These processors thin the data, perform feature extraction, etc. The problem is that even with favorable projections of advances in electronics, the size, weight, and power of electronic real time processors cannot be supported by many multispectral imaging system platforms. There is a need to develop a compact, very low power processor with substantial processing capability to meet existing requirements for real time on-board processing of multispectral imaging system data streams. More specifically, there is a need for a real time processor capable of correlating an input spectrum against template spectra, searching for spatial scene pixels that contain the same spectrum as a designated spectrum or spatial pixel, incorporating atmospheric and calibration corrections, as well as performing other functions.
The improved multispectral imaging system and method of the present invention address these needs by exploiting the three-dimensional attributes of optical correlation to perform massively parallel correlation processing. The expression xe2x80x9cmultispectral imagingxe2x80x9d as used herein is intended to encompass systems having several spectral bands, hundreds of bands or even finer spectral xe2x80x9cbinningxe2x80x9d, as referred to above.
The multispectral imaging system according to the invention comprises a multispectral imager for producing spatial pictures of an image scene in a plurality of different spectral wavelength bands. The system further includes an optical processor for simultaneously comparing an input wavelength spectrum observed in a scene image from the multispectral imager with a plurality of template wavelength spectra to find a correlation.
The optical processor includes a first spatial light modulator having an array of modulating elements including n modulating elements arranged linearly for receiving respective ones of n spectral bands of an input wavelength spectrum observed in a scene image from the multispectral imager and independently altering at least one property of an incident light beam by a value corresponding to the observed intensity of the input spectrum in the respective spectral band. A means is provided for expanding a modulated beam of an input wavelength spectrum from the n elements of the first spatial light modulator in a spatial dimension perpendicular to the linear direction of the n elements of the modulator.
A second spatial light modulator of the system in a disclosed embodiment has a two-dimensional array of nxc3x97m modulating elements through which the expanded, modulated beam of an input wavelength spectrum from the n elements of the first spatial light modulator is transited. Each row n of the nxc3x97m array of elements alters the at least one property of the incident light by values corresponding to a particular template wavelength spectrum of m such template wavelength spectra of the two-dimensional array of elements of the second spatial light modulator. These values corresponding to each template spectrum of the m template spectra are the conjugates of the representative values of the n elements of the template spectrum of the m template spectra.
The at least one property of the incident light of the input spectrum provided to the first spatial light modulator is preferably uniform. Thus, where the input spectrum corresponds to a particular template spectrum of the m template spectra of the second spatial light modulator, the at least one property of the twice modulated light is first altered and then changed back to uniform by the system to permit focusing the light from the second modulator and detecting the intensity thereof to indicate a correlation in the case the input spectrum corresponds to the particular template spectrum of the m template spectra.
The related method of spectral imaging of the present invention utilizes this optical processing of the invention for simultaneously comparing a wavelength spectrum observed for a spatial pixel of a spatial picture of an image scene with a plurality of template wavelength spectra to find a correlation. The simultaneously comparing can be performed simultaneously for each of a plurality of different spatial pixels in the scene image using a two-dimensional spatial light modulator as the input device, rather than a linear spatial light modulator.